Visual-Semantic-Encoder/train.py at master · RishitModi/Visual-Semantic-Encoder · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
import torch
from torch.nn.utils.rnn import pad_sequence
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torch.utils.data import Dataset, DataLoader, Subset
import torchvision.transforms as transforms
import os
import json
from models import combined_model
from utils import dataset

def save_checkpoint(model, optimizer, epoch, loss, path="checkpoint_img_caption.pth"):
    torch.save({
        'epoch': epoch,
        'model_state_dict': model.state_dict(),
        'optimizer_state_dict': optimizer.state_dict(),
        'loss': loss
    }, path)


def load_checkpoint(model, optimizer, path="checkpoint.pth"):
    checkpoint = torch.load(path, map_location="cpu")
    model.load_state_dict(checkpoint['model_state_dict'])
    optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
    return checkpoint['epoch'], checkpoint['loss']

from tqdm import tqdm
import matplotlib.pyplot as plt

def train_model(model, dataloader, criterion, optimizer, vocab, device, num_epochs=1):
    model.to(device)
    train_losses = []

    for epoch in range(num_epochs):
        model.train()
        total_loss = 0

        loop = tqdm(dataloader, desc=f"Epoch [{epoch+1}/{num_epochs}]", leave=False)
        for images, captions in loop:
            images = images.to(device)
            captions = captions.to(device)

            outputs = model(images, captions[:, :-1])
            targets = captions[:, 1:]

            # Handle possible mismatch
            min_len = min(outputs.size(1), targets.size(1))
            outputs = outputs[:, :min_len, :]
            targets = targets[:, :min_len]

            outputs = outputs.reshape(-1, outputs.shape[2])
            targets = targets.reshape(-1)

            optimizer.zero_grad()
            loss = criterion(outputs, targets)
            loss.backward()
            optimizer.step()
            total_loss += loss.item()
            loop.set_postfix(loss=loss.item())

        avg_loss = total_loss / len(dataloader)
        train_losses.append(avg_loss)
        print(f"Epoch [{epoch+1}/{num_epochs}] Avg Loss: {avg_loss:.4f}")

        save_checkpoint(model, optimizer, epoch, avg_loss)

    plt.figure(figsize=(8, 5))
    plt.plot(range(1, num_epochs+1), train_losses, marker='o', color='blue')
    plt.title("Training Loss over Epochs")
    plt.xlabel("Epoch")
    plt.ylabel("Loss")
    plt.grid(True)
    plt.show()

    return train_losses

import matplotlib.pyplot as plt
from PIL import Image
import cv2
import random
import torch

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Define paths and prepare dataset
datasetpath = "/kaggle/input/coco-2017-dataset/coco2017"
trainimagepath = os.path.join(datasetpath, "train2017")
annotationspath = os.path.join(datasetpath, "annotations")
captionsjsonpath = os.path.join(annotationspath, "captions_train2017.json")

with open(captionsjsonpath, "r") as f:
    captionsjson = json.load(f)

idtofilename = {img["id"]: img["file_name"] for img in captionsjson["images"]}

data_pairs = []
for ann in captionsjson["annotations"]:
    img_id = ann["image_id"]
    if img_id in idtofilename:
        data_pairs.append((os.path.join(trainimagepath, idtofilename[img_id]), ann["caption"]))

all_captions = [cap for _, cap in data_pairs]
vocab = Vocabulary(freqthreshold=5)
vocab.buildvocabulary(all_captions)

transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    )
])

dataset = CocoDataset(data_pairs, vocab, transform=transform)
subset_dataset = Subset(dataset, list(range(60000)))

dataloader = DataLoader(
    subset_dataset,
    batch_size=32,
    shuffle=True,
    collate_fn=collate_fn
)

embed_size = 256
hidden_size = 512
vocab_size = len(vocab)

model = CombinedModel(embed_size, hidden_size, vocab_size).to(device)

criterion = nn.CrossEntropyLoss(ignore_index=vocab.stoi["<PAD>"])
optimizer = optim.Adam(model.parameters(), lr=1e-3)

# Train
train_model(model, dataloader, criterion, optimizer, vocab, device, num_epochs=60)